Zero-shot Medical Entity Retrieval without Annotation: Learning From Rich Knowledge Graph Semantics

TL;DR

This paper introduces a knowledge graph-based approach for zero-shot medical entity retrieval that outperforms existing benchmarks without requiring human annotations, addressing the challenge of generalizing to unseen medical sub-specialties.

Contribution

It proposes a novel learning framework leveraging medical knowledge graphs to enable zero-shot retrieval without human annotation, improving recall across major medical ontologies.

Findings

Outperforms BM25 and Clinical BERT by 7-30% in recall

Effective across UMLS, SNOMED, and ICD-10 ontologies

Does not require human annotation for training

Abstract

Medical entity retrieval is an integral component for understanding and communicating information across various health systems. Current approaches tend to work well on specific medical domains but generalize poorly to unseen sub-specialties. This is of increasing concern under a public health crisis as new medical conditions and drug treatments come to light frequently. Zero-shot retrieval is challenging due to the high degree of ambiguity and variability in medical corpora, making it difficult to build an accurate similarity measure between mentions and concepts. Medical knowledge graphs (KG), however, contain rich semantics including large numbers of synonyms as well as its curated graphical structures. To take advantage of this valuable information, we propose a suite of learning tasks designed for training efficient zero-shot entity retrieval models. Without requiring any human annotation, our knowledge graph enriched architecture significantly outperforms common zero-shot benchmarks including BM25 and Clinical BERT with 7% to 30% higher recall across multiple major medical ontologies, such as UMLS, SNOMED, and ICD-10.